using AI, meta generates videos from text with ‘make-a-video’

Meta introduces AI make-a-video

Facebook’s Meta calls its new artificial intelligence system Make-A-Video state-of-the-art. It builds on the recent progress that text-to-image generation technology has undergone which uses descriptions to churn out moving images from the texts. Meta has taken the advancement under its wing as it joins the troupe and lets its audience create their own GIF or videos with just a few words or lines of text.

Some of the descriptions the team has tried include ‘A dog wearing a Superhero outfit with a red cape flying through the sky,’ ‘Cat watching TV with a remote in hand,’ ‘An oil painting of a couple in formal evening wear going home get caught in a heavy downpour with umbrellas’, and ‘A fluffy baby sloth with an orange knitted hat trying to figure out a laptop close up highly detailed studio lighting screen reflecting in its eye.’ The first two seem to have generated the exact images with just a few words while the latter two, albeit releasing exact images as well, required more details. Still, the moving images produced captured the descriptions from the team’s texts.

images courtesy of Meta AI

More internet content for creators and artists

Meta says that Make-A-Video opens opportunities for content creators and artists by offering them tools to quickly and easily create new content through generative AI. ‘The system learns what the world looks like from paired text-image data and how the world moves from video footage with no associated text,’ writes Meta. The system can also create videos from images or take existing videos and create new ones that are similar. Take, for example, the header image of this article where the left one is the existing image while the right one is what Make-A-Video generated, a moving image edited with a myriad of face-enhancing effects. ‘We want to be thoughtful about how we build new generative AI systems like this,’ adds Meta.

a dog wearing a superhero outfit with red cape flying through the sky

Publicly available datasets and research

Users can also upload a static image, and Make-A-Video from Meta will transform it into a moving image, thanks to the generative AI technology. Make-A-Video uses publicly available datasets, and Meta is openly sharing its generative AI research and results with its community for their feedback. It says it will continue to use ‘our responsible AI framework to refine and evolve our approach to this emerging technology.’ The source data of Make-A-Video analyzes millions of pieces of data to learn about the world. To reduce the risk of harmful content being generated, the team examines, applies, and iterates on filters to reduce the risk of harmful content surfacing in videos. Meta adds its watermark in every video it produces for a reason. ‘This will help ensure viewers know the video was generated with AI and is not a captured video,’ writes Meta.

cat watching TV with a remote in hand